The Smart Factory Signal: It’s Not Token Burn, It’s Where the Tokens Burn

By Wim Dijkgraaf
CEO and Founder at Quotation Factory and HiggX

For a while, I thought token burn might become one of the best indicators of whether an organisation is actually becoming intelligent.

It is a tempting idea.

In a world full of AI strategy decks, innovation workshops, pilots, and vague transformation language, token consumption looks refreshingly concrete. It is measurable. It is real. It is hard to fake at scale. If an organisation is burning a serious volume of tokens, something is happening.

But the more I think about it, the more I believe token burn is only useful if we ask a more important question:

Where are the tokens being burned?

That distinction matters more than the total number.

Because not all token burn means the same thing. In fact, one kind of token burn may tell us very little about whether a company is becoming smarter, while another kind may be one of the clearest early signals that intelligence is being embedded into the operation itself.

And that, to me, is where the smart factory discussion gets interesting.

Every factory has two organisations inside it

To understand this, it helps to look at a factory as if it contains two different organisations.

The first is the change organisation.

This is the part of the business that works on transformation. It runs innovation programmes, process redesign, digital initiatives, pilots, workshops, proofs of concept, data projects, and AI experiments. It is where a company imagines its future and tries to move toward it.

The second is the going concern.

This is the real operating system of the company. It is the daily machinery of planning, production, procurement, quality, engineering response, maintenance, customer delivery, scheduling, issue handling, and all the small and large decisions that keep the business moving.

Both matter. Both consume energy. Both can use AI.

But they should not be interpreted in the same way.

In the change organisation, token burn measures effort

When tokens are being burned in the change organisation, that usually tells us something useful — but limited.

It tells us the company is exploring. It tells us people are testing, learning, experimenting, and investing effort into innovation. That is not trivial. It matters. Most organisations do not change by accident.

If transformation teams are using AI heavily, they are probably generating ideas faster, testing scenarios more broadly, producing more options, and accelerating redesign work. That can be valuable. It may even be necessary.

But it does not yet mean the factory itself is smart.

It may simply mean the organisation is trying to become smarter.

That is an important difference.

A company can burn an enormous amount of tokens in strategy sessions, process workshops, copilots, demo environments, internal presentations, innovation labs, and pilot projects — and still have almost no embedded intelligence in the daily operation.

That is why token burn in the change organisation is best interpreted as a sign of ambition.

It measures effort. It measures intent. It measures transformation activity.

But it does not yet measure operational intelligence.

The real question is whether tokens have crossed into the going concern

The signal becomes much more meaningful when token consumption shows up in the going concern.

That is where AI stops being a topic and starts becoming infrastructure.

If planners are using AI every day to resolve scheduling conflicts, if quality teams are using it to accelerate root-cause analysis, if maintenance teams are using it to interpret failure patterns, if operators are receiving contextual support in the flow of work, if engineering teams are translating customer and production changes through AI-assisted workflows — then something fundamentally different is happening.

At that point, AI is no longer sitting on the edge of the organisation as a transformation project.

It has crossed the boundary into execution.

And that is a far more revealing sign of a smart factory than how many pilots are running or how many innovation presentations have been delivered.

A smart factory is not one that merely discusses intelligence. It is one that uses intelligence in the ordinary flow of work.

That is the threshold that matters.

Because the factory does not become smart when AI appears in a roadmap. It becomes smart when AI starts participating in the daily metabolism of the business.

From experimentation to absorption

This is why I believe the deeper issue is not token usage, but organisational absorption.

A lot of companies are currently in the experimentation phase. That is natural. They are testing tools, training teams, building prototypes, and trying to understand where the value might be. In that phase, token burn can rise quickly without changing much in the actual operation.

That is not failure. It is transition.

But the real transformation only begins when intelligence is absorbed into the routines of the enterprise.

Not occasionally. Not symbolically. Not as a showcase.

Structurally.

When intelligence becomes part of how production is planned, how exceptions are handled, how knowledge is retrieved, how changes are processed, how service decisions are made, and how operations improve themselves, then the company is no longer experimenting with AI. It is reorganising work around it.

That is what makes a factory smart.

Not possession of advanced tools, but absorption of intelligence into execution.

That is also why many highly visible AI programmes are less impressive than they first appear. They often show that an organisation has innovation capacity, not necessarily operational transformation.

The real test is quieter.

It shows up when daily work changes.

Beware of confusing activity with intelligence

This is also where some caution is needed.

Even token burn in the going concern is not automatically proof of smartness.

A factory can consume large volumes of tokens in operations and still use AI badly.

It could be solving low-value problems. It could be compensating for poor process design. It could be creating expensive dependencies for tasks that should have been standardized. It could be duplicating decision support. It could be generating activity without creating leverage.

In other words, more token burn does not necessarily mean more intelligence. It may simply mean more consumption.

That is why token burn should be treated as a proxy, not as the KPI itself.

It is useful because it can reveal where intelligence is being used. But it does not tell us whether that intelligence is creating value unless it is connected to real operational outcomes.

So the smart factory is not the one that burns the most tokens.

It is the one that creates the most value from intelligence embedded in the work.

A better way to read the signal

If we want to use token burn at all, I think we should read it as part of a simple maturity model.

The first level is innovation burn.

This is when tokens are mostly consumed by transformation teams, digital programmes, AI pilots, consultants, or redesign initiatives. The organisation is exploring. That is good. But it is still early.

The second level is operational burn.

This is when token consumption starts appearing inside the daily workflows of the going concern. Planning, quality, maintenance, engineering, production support, customer operations — this is where the signal becomes serious. Intelligence is crossing into real execution.

The third level is productive burn.

This is when operational token use is clearly associated with measurable performance improvements. Lead times improve. Engineering response accelerates. Quality issues are resolved faster. Knowledge becomes more available. Rework is reduced. Decisions become better and faster. The organisation is not just using intelligence; it is compounding value from it.

That is the point at which AI has moved beyond novelty and into industrial relevance.

The smart factory is not the one with the loudest AI programme

This matters because too many organisations still confuse AI visibility with AI maturity.

It is easy to celebrate the visible things: the innovation team, the shiny pilot, the demo, the internal chatbot, the strategic partnership, the presentation at the executive offsite. Those things are legible. They can be named, funded, and announced.

Operational embedding is much harder.

It is messier, slower, more political, and less glamorous. It touches real workflows, real roles, real accountabilities, real systems, and real habits. It forces organisations to redesign not just what they know, but how they work.

That is why the most intelligent factories may not always be the ones talking most loudly about AI.

They may simply be the ones where intelligence has quietly entered planning, engineering, quality, service, and execution.

The ones where people no longer ask whether AI is being used, because it has become part of the operating environment.

That is the shift leaders should be watching.

Not just: How much AI are we using?

But more importantly: Where has intelligence actually been absorbed into the business?

Don’t measure the theatre. Measure the metabolism.

So, is token burn the best indicator of a smart factory?

Not exactly.

On its own, it is too crude. Too easy to misread. Too easy to optimize for the wrong reasons.

But the location of token burn may be one of the most revealing early signals we have.

If the tokens are mostly being consumed in the change organisation, that tells us the company is investing in innovation. That is worthwhile, but it is not yet proof of operational intelligence.

If the tokens are increasingly being consumed in the going concern, that is more interesting. It suggests intelligence is moving into the real workflows of the enterprise. It suggests AI is no longer an experiment on the side, but a participant in daily execution.

And if that operational use is tied to better performance, better responsiveness, better quality, and better decisions, then we are no longer looking at AI theatre.

We are looking at the beginnings of a smart factory.

That is the distinction I would pay attention to.

Not whether an organisation is burning tokens.

But whether it is burning them where value is actually created.